DREAM Challenge Overview
Attila Gabor and Marco Tognetti
3/20/2019
Overview : challenge 1
Challange 1 is about predicting the time courses of some markers in some cell-lines.
Cell-lines show various trends in response to stimuli and certain drug treatments.
We have divided the measured cell-lines to a trainig and a test and a validation set.
In the following section we show some examples of these features from the training set.
Your task will be to predict the interestingly behaving markers’ time courses in the test and validation set.
Input data shaping:
## Warning: package 'pheatmap' was built under R version 3.5.2## # A tibble: 130,118 x 7
## cell_line treatment time time_course cellcount reporter value
## <fct> <fct> <dbl> <fct> <int> <chr> <dbl>
## 1 HCC1428 iPI3K 9 A 8831 IdU 2.60
## 2 HCC1428 EGF 23 A 7749 IdU 2.61
## 3 HCC1428 iPI3K 40 A 8010 IdU 2.59
## 4 HCC1428 imTOR 40 A 8465 IdU 2.59
## 5 HCC1428 iEGFR 40 A 8023 IdU 2.45
## 6 HCC1428 iPKC 40 A 6395 IdU 2.51
## 7 HCC1428 iMEK 40 A 7289 IdU 3.27
## 8 HCC1428 imTOR 9 A 10871 IdU 2.57
## 9 HCC1428 iEGFR 9 A 9942 IdU 2.50
## 10 HCC1428 imTOR 13 A 8559 IdU 2.62
## # … with 130,108 more rowsShow Interesting features to support challenge questions
Response to EGF
on absolute scale
pdf("./figures/median_phospho_EGF_response.pdf")
for(marker in unique(phospho_median$reporter)){
gg = phospho_median %>%
filter(treatment=="EGF", reporter == marker) %>%
group_by(cell_line, time) %>% summarise(value=mean(value)) %>%
ggplot(aes(time,value)) +
geom_line() +
facet_wrap(~cell_line) + theme_bw() + guides(color=FALSE) + ggtitle(marker)
print(gg)
}
dev.off()## quartz_off_screen
## 2pCREB
creb_responders = c("HCC1428", "HCC1569","HCC1599", "HCC2218","MDAMB468")
creb_nonresponders = c("MDAkb2","MACLS2", "T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == "p-CREB") %>%
filter(cell_line %in% c(creb_responders,creb_nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle("p-CREB",subtitle = "responding and non-responding cell-lines")
pSTAT5
pSite = "p-STAT5"
responders = c("AU565", "HCC1806","HCC1187", "MCF12A","MDAMB468")
nonresponders = c("BT20","BT483", "T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
# pSRC I dont see cell lines responding to EGF
p-FAK
mostly non-responsive
pSite = "p-FAK"
responders = c("DU4475", "MCF12A","BT20")
nonresponders = c("UACC893","KPL1", "HCC38")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
## p-MEK
pSite = "p-MEK"
responders = c("AU565", "DU4475","HCC2157","MDAMB436")
nonresponders = c("BT483","KPL1", "MX1","ZR7530")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
p−S6K
pSite = "p-S6K"
responders = c("AU565", "DU4475","BT483","HCC1187")
nonresponders = c("BT483","KPL1", "CAL120","MFM223","MPE600")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
p−MAP2K3
pretty noisy and the range of values are narrow
p−STAT1
pSite = "p-STAT1"
responders = c("AU565", "HCC1806", "BT20","MDAMB468")
nonresponders = c("BT483", "T47D","HCC1187")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
## p−p53 does not have much dynamics
pSite = "p-p53"
responders = c("184B5", "DU4475", "HCC1143","Hs578T")
nonresponders = c("BT483", "T47D","HCC1187")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
p−NFkB
pSite = "p-NFkB"
responders = c("184B5","AU565" ,"BT549", "CAL851","HCC1937","MCF12A","HCC1954")
nonresponders = c("BT483", "T47D","EFM19")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responding and non-responding cell-lines")
## p−p38 we nice dynamics !!
pSite = "p-p38"
responders = c("184B5","AU565" ,"BT549", "CAL851","HCC1937","MCF12A","HCC1954","BT483", "T47D")
nonresponders = c("EFM19")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "plateau and peaks")
## p−AMPK increasing and decreaseing dynamics
pSite = "p-AMPK"
increaseing = c("184A1","BT20", "HCC1937", "HCC1937","HCC1937")
decreasing = c("CAL851", "HDQP1","MCF10F","ZR751","HDQP1")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(increaseing,decreasing)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "up and downs")
p-AKT_S473
increasing and decreaseing dynamics
pSite = "p-Akt(Ser473)"
plateau = c("AU565","HBL100", "HCC1937", "HCC70","Hs578T")
peaks = c("CAL851", "MCF12A","MDAkb2","UACC893","SKBR3")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(plateau,peaks)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "up and downs")
## p-ERK increasing and decreaseing dynamics
pSite = "p-ERK"
plateau = c("AU565", "MDAMB436", "HCC1395","CAL51")
peaks = c("HBL100", "MCF12A","MDAkb2","UACC893","SKBR3")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(plateau,peaks)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "up and downs")
CyclinB
no much dynamics
GSK3b
pSite = "p-GSK3b"
responders = c("184B5", "CAL120", "AU565","MDAMB157","UACC893")
nonresponders = c("HBL100", "HDQP1","MDAkb2","T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
GAPDH
no strong response
p-MKK3-MKK6
pSite = "p-MKK3-MKK6"
responders = c("184A1", "HBL100", "CAL120", "HCC2185","MDAMB157","HCC70")
nonresponders = c("AU565", "HDQP1","MDAkb2","T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
### p-PDPK1 effect is not really clear
pSite = "p-PDPK1"
responders = c("184B5", "MX1")
nonresponders = c("HCC1806","T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
### p-BTK effect is not really clear
pSite = "p-BTK"
responders = c("HCC2185","DU4475", "HCC2185","MCF7","ZR7530","184A1")
nonresponders = c("MDAkb2","T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
p-p90RSK
good signals
pSite = "p-p90RSK"
plateau = c("AU565", "MDAMB436", "HCC1395","CAL51")
peaks = c("HBL100", "MCF12A","MDAkb2","UACC893","SKBR3")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(plateau,peaks)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
### p-SMAD23 effect is not really clear
pSite = "p-SMAD23"
samples = c("DU4475", "CAL120", "HCC70","T47D")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(samples)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
p-STAT3
clearly most of the time no response, sometimes good peaks.
pSite = "p-STAT3"
responders = c("184A1", "184B5", "BT20","HCC1806","MDAMB468")
nonresponders = c("HCC2218","MPE600")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
p-JNK
effect is not really clear
pSite = "p-JNK"
responders = c("184A1", "184B5", "BT549","MCF10A","SKBR3")
nonresponders = c("HCC1806","MDAMB453")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
### Ki-67 not much dynamics
p-H3
effect is not really clear
pSite = "p-H3"
responders = c("184A1", "184B5", "HCC1937","MCF10A","SKBR3","UACC893")
nonresponders = c("BT549","T47D","DU4475")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
### p-S6 really nice signals
pSite = "p-S6"
responders = c("184A1", "HCC1187", "MCF12A","MCF10A","SKBR3","UACC893")
nonresponders = c("CAL51","EFM192A","MACLS2")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
cleavedCas
not clear.
p-MKK4
nice signals ! plateau and peaks
pSite = "p-MKK4"
responders = c("AU565", "HCC2218", "CAL120","HCC1187","T47D","MDAMB157")
nonresponders = c("MCF7","HCC1500","HCC70")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
p−AKT(Thr308)
many unclear but some are nice .
pSite = "p-AKT(Thr308)"
responders = c("MCF12A", "HCC1143", "DU4475","Hs578T","T47D")
nonresponders = c("MCF7","HCC1500","HCC70")
phospho_median %>%
filter(treatment=="EGF") %>%
filter(reporter == pSite) %>%
filter(cell_line %in% c(responders,nonresponders)) %>%
ggplot(aes(time,value,group=cell_line)) +
geom_line() +
facet_wrap(~cell_line) +
theme_bw() +
guides(color=FALSE) +
geom_smooth(formula = "y~x", method = "loess") +
ggtitle(pSite,subtitle = "responders and non-responsive nodes")
4EBP1 and RB
not much dynamics